Wear, 74 (1981-
27 - 50
FRETTING CORROSION OF GOLD-PLATED CONNECTOR CONTACTS*
and M. H. DROZDOWICZ
Bell La~~ruto~es, Columbus, OH 43213 (US.A.J (Received
April 6, 1981)
Summary Base metal contacts in separable electronic connectors that are subjected to vibration, mechanical shock or thermal fluctuations often develop a high contact resistance due to fretting corrosion. The degradation of contact behavior is due to the accumulation of insulating oxides at the interface. Gold-plated contacts may also fail by fretting if the finish is worn through. A study of base and electroplated gold contacts with the gold in a range of thicknesses revealed that transfer dominates fretting and that an equilibrium distribution of a contact metal with an underplate and a substrate is attained which is the same for both members, although they may initially have been different. A thin gold deposit on a nickel underplate maintains a low contact resistance when the opposing member is solid gold, because nickel promotes adhesive transfer of the gold. However, solid 7OAu-30Ag alloy transfers less well, thereby destabilizing the contact resistance of the system. Lubrication is of benefit as it lowers the gold wear rate, which delays the appearance of base material. Solder plate readily transfers to a mating gold contact, thus making both surfaces base, with rapid escalation of the contact resistance.
1. Introduction The contact resistance of separable electronic connectors must be low and stable, despite the mechanical wear, environmental stresses and other degrading influences to which they may be subjected for lifetimes of as much as 40 years in some telecommunications applications. The most reliable connectors for low voltage, low current circuits usually have contacts made of gold or high gold alloys. Gold does not form an insulating oxide, a tarnish or corrosion films; it has other desirable properties such as a high
*Paper presented at the International Conference Francisco, CA, U.S.A., March 30 - April 1, 1981. 0 Elsevier
on Wear of Materials,
in The Ne~erl~ds
conductivity, a high durability and a high solderability; in addition it is easily fabricated by electroplating, cladding and welding. Well over lo6 oz tr of gold are used annually for electronic connectors and the edge contacts of printed-circuit boards. Rapid escalation of the price of gold has stimulated a re-evaluation of thickness requirements. Where 0.5 - 5 pm of gold formerly was the standard [l] , depending on the connector design and application, 0.25 - 2.5 ym is the range now commonly used. Degradation processes become more likely when thinner gold films are employed. One of these potential failure mechanisms is fretting corrosion. Fretting corrosion is a highly localized wear phenomenon in which debris accumulates at the interface between contacting members undergoing cyclic or repetitive small-amplitude relative motion. Such movements may occur when a connector is subjected to vibration, mechanical shock or thermal excursions. The wear debris from base metals contains products of the reaction with the environment, such as oxides. Thus, fretting corrosion can cause substantial increases in the contact resistance. Fretting corrosion is considered  to be a significant failure mechanism for electronic connectors which have tin- and solder-plated contacts. Unexplained rises in the contact resistance that have been observed [ 31 when mated connectors with thin gold platings were aged were conceivably due to fretting corrosion if the deposit wore through to the base metal. No systematic studies of fretting corrosion with gold-plated contacts have been reported. The objective of this investigation was therefore to determine the behavior of gold plate, and in particular to learn how the contact resistance stability depends on the thickness of the finish. Other factors which might control fretting corrosion, such as contact lubricants and underplatings, were explored. Fretting corrosion was modeled with an apparatus consisting of a rider on a flat specimen; this permitted the operational parameters and contact materials to be varied. Riders of solid metals and plated flats were used. It was found that the stability of the contact resistance depends, as expected, on maintaining the Au-Au contact and that thick deposits are more desirable than thin deposits. Contact lubrication is beneficial because it lowers the wear rate of the gold. With the combinations of metals used, an equilibrium distribution of gold adhesive transfer fragments with underplate or substrate metals is achieved after prolonged fretting which is the same for both members. In turn, the composition of this mixture controls the contact resistance. The nickel underplate is wear resistant and tends to maintain a low contact resistance by promoting transfer from the opposing gold contact. Solid 7OAu-30Ag alloy is less favorable because it wears less than pure gold. Thus, once the gold finish has worn away, the mixture of materials at the interface is gold poor with a resulting increase in the contact resistance. A very soft base metal, electroplated 60Sn-40Pb, readily transfers to its mating gold contact, which makes both surfaces base with a consequent rapid escalation in the contact resistance.
Fig. 1. A block diagram of the fretting apparatus.
2. Experimental details 2.1. Apparatus Connector contacts were simulated by using configuration of a rider on a flat specimen. The stationary rider, mounted on a ftexure plate, is dead weight loaded against a moving flat. Both members are made from actual or experimental contact materials. The flat, on a high resolution positioning table, is driven by a d.c. stepping motor through a micrometer screw 2.5 cm long and with 16 turns cm‘-‘. The motor is interfaced to a computer which controls the speed of the table and track length with a resolution of 1.6 I.tm. At the end of a run the parts are separated with a spring gauge so as to measure the adhesive force if cold welding occurred. The entire apparatus is mounted on an air table for isolation from external vibrations. Figure 1 further illustrates the operation of the apparatus. The operator gives commands to a data acquisition system (DAS) computer via a teletypewriter (TTY) as to the velocity, track length and number of cycles in a run. This information is transmitted by the computer to the translator module which provides sequencing and switching logic for bidirectional control of the stepping motor. The velocity is controlled within the range 1.6 - 320 E.tm s-l in increments of 1.6 pm for track lengths of 1.6 pm - 13 mm. The stepping motor moves in discrete identical steps. Because of this, the motion of the table can be intermittent or nearly continuous depending on how fast the motor is pulsed. In most of the experiments, movement was at 100 pulses (160 pm) s-l except when the contact resistance was being determined, and sliding was essentially continuous in each direction. The contact resistance was determined using a d.c. four-wire dry circuit technique in accordance with ref. 4. The circuit has an opencircuit voltage of 20 mV with current limited to 100 mA. These conditions were not expected to cause any physical changes in the contact junction, such as the breakdown of thin insulating films or the softening of contact asperities which might affect fretting. This was verified by conducting typical runs both ‘with and without a current, and it was found by optical examination that the worn surfaces were identical. When current is passed through the junction, the voltage drop is monitored by a programmable digital voltmeter (DVM) through an operational amplifier (OP-AMP). The OP-AMP provides for zeroing and an increase
RUN 875 . ..SOLID AU RIDER VS 0.6UM AU OVER CU FLAT SAMPLE t6632B.e. 2OUM TRACK LENGTH . ..159UM/SEC VELOCITY 33% RH... 23 DEGREES C . ..TIME 2r19 PM TRACK
CONTACT RESISTANCE FOR CYCLE # 1 .70 -70 .70 .69 .68 .66 .64 .63 -61 .59 .57 .55 .54 .59 CONTACT RESISTANCE FOR CYCLE # 10 -36 .36 .37 .37 .37 .37 .37 .36 .32 .33 .33 .33 .33 .37 CONTACT RESISTANCE FUR CYCLE # 20 .32 .32 .33 .33 .34 .34 .34 .33 ,33 .33 .32 -32 .28 .32
CONTACT .28 .28 RESISTANCE .28 .27 .28 FOR -27 CYCLE .27 .27 # .26 5:8 .27 .27 ‘28 .29
Fig. 2. A computer print-out of the contact resistance from a typical run.
in the signal of a factor of 10. The DAS samples the contact resistance at a preprogrammed number of cycles. At these times the table is stopped at the beginning of a wear track and the DAS takes four contact resistance measurements and stores them. The table then moves one step and another four measurements are made. This continues for the length of the track. Each set of me~~ments is averaged by the computer and printed out on the TTY, as shown in Fig. 2. Contact resistance data are presented as plots of the maximum values from the print-out against number of cycles of fretting. 2.2. Materials The contact materials used are given in Tables 1 and 2. 2.2.1. Riders The riders were smooth hemispherically ended rods of diameter 3.2 mm. Most runs were with solid pure gold or solid 70Au-30Ag ahoy, materials which are commonly used in connectors. Runs were also made with riders of the base metals, copper, Be-Cu alloy, nickel and solder-plated copper, to obtain comparative data. A new rider was used for each run. 2.2.2. Flats Copper flats, 3.8 cm X 1.3 cm X 0.32 cm, were randomly abraded by hand on metallographic papers to several roughnesses. They were then plated with a cobalt-hardened gold in thicknesses from 0.05 to 0.6 pm, both with and without 2.5 pm of nickel underplate. These thicknesses were estimated gravimetrically to within 0.01 ,um from the projected areas of the surfaces and densities of the deposits. Cobalt-hardened gold electroplate is the most widely used contact finish in the connector industry. Base metal flats were also employed in some runs. Immediately prior to a run, specimens were cleaned in multiple baths of warm reagent-grade l,~,l-t~chloroe~~e and methanol.
31 TABLE 1 Solid contact materials Material
Au rider Au alloy rider Cu, oxygen-frees rider, flat Be-Cu alloya rider, flat Ni rider
99.99 7OAu-30Ag 99.95 97.9 Cu(1.9 Be, 0.2 Ni) 99.4 (Ni + Co)
Hardness (typical) (kgf mmm2) 65 100 105 - 135 215 - 230 250
aRoughnesses of fiats: Cu, 0.4 pm c.1.a.; Be-C& 0.2 pm c.1.a.
TABLE 2 Electrodeposited material
Au flat Ni flat Sn-Pb rider Sn-Pb flat
contact materials ~~derpZate (Pm)
0.05, 0.13, None, Ni 0.23, 0.6 (2.5) 2.5 25 25 Cu(75) electrolytic foil
Hard~e~a (kgf mm-‘)
Cud cu Epoxy-Dacron printed-circuit board
99 +e 60Sn-40Pbf 60Sn-40Pbf
450 20 20
aHardness determined on a metallographic section of thick deposits; Knoop; typical values. bRoughness after plating, 0.1, 0.4, 0.9 pm c.1.a. ‘Gold plated from a cobalt-doped citrate-buffered solution of KAu(CN)g ; typical composition, 0.2% Co, 0.2% K, 0.2% C. dRoughness after plating, 0.2 pm c.1.a. eNickelplated from a Watts bath. fSolder plated from a fluoborate bath.
2.2.3. Lubrication The lubricant was a liquid polyphenyl ether , OS-124 (registered trademark of the Monsanto Chemical Company, St. Louis, MI), commonly used on connectors. It was applied to the cleaned flats by immersion and with~aw~ at room temperature from a 0.5% solution in l,~,l-tri~hloroethane. The solvent quickly evaporated, leaving a thin residual film of lubricant. 2.3. Selection of test conditions In planning the experiments it is necessary to define the fretting amplitude, the cycle rate, the load, the test duration and the environment. Measurements of the amplitude of slip do not appear to have been made for con-
nectars. For vibration-induced movements it will depend on the mass, the acceleration, the friction at the interface and the compliance of the parts; some connectors are rigid, while others fret easily. For the differential thermal expansions and contractions of the members to which the mating contacts are attached, amplitudes from a few to several tens of microns may occur. In a calculation of a typical case, if we assume that an edge connector and a glass epoxy printed-circuit board 15 cm long are rigidly mounted to opposite sides of a steel frame, a 20 “C fluctuation will produce a wipe distance as large as 56 pm, depending on the board material (the coefficients of thermal expansion are in the range [ 61 (11 - 40) X 10m6 “C-i; for steel the thermal expansion coefficient is 12 X 10e6 “C-l). The value, 20 pm, was chosen as a realistic wipe distance for this study. The contact resistance instability of base metal contacts has been found to diminish with increasing load  . Runs with the noble metals of this study, discussed later, often resulted in cold welding in the absence of lubrication, a condition which became more severe as the load was increased. A relatively small load of 50 gf was therefore chosen; this contact force is near the low end of the range used in electronic connectors. Runs with gold contacts were continued usually for 300 000 or 1 000 000 cycles (a to-and-fro movement is 1 cycle). This was sufficient to produce wear out of the gold plating and an equilibrium condition of contact resistance and surface composition in most cases. Runs with all-base metal contacts were much shorter, usually terminating when the contact resistance rose to 1000 Sz from its initial value of several milliohms. The changes in the contact resistance were found to depend on the cycle rate, as discussed below, and these appeared after fewer cycles but a longer real time with a diminishing cycle rate. Accordingly, a standard rate of 4 Hz (100 pulses s-i of the stepping motor) was used. This rate was experimentally convenient and had a significant effect on the contact resistance from relatively short experiments. The environment was not controlled. However, all runs were in laboratory air at 20 - 25 “C! and 20% - 50% relative humidity.
3.1. Effect of the cycle rate on the contact resistance Runs at 100 gf were conducted with solid copper riders and flats, and with solid gold riders mated to Co-Au plate 0.05 pm thick on copper. The cycle rate ranged from 0.04 to 2 Hz. The gold plate wore out during the runs. Plots of the contact resistance against the number of cycles are given in Fig. 3. Both materials systems show that, the slower the rate, the smaller the number of cycles required to attain a given increase in contact resistance. This effect is probably due to the kinetic factor of oxidation; the lower the
Fig. 3. Effect of the cycle rate on the contact resistance: A, 0, n, copper rider against copper; 0.0, solid gold riders against 0.05 pm Co-Au on copper.
frequency, the more time is available for the oxidation of copper exposed by fretting wear which, in turn, is responsible for the increase in contact resistance. The dependence of the increase in contact resistance on the cycle rate is similar to the weight loss-frequency relationships reported in ref. 7, pp. 115,235, for ferrous systems; frequency effects were not observed above about 15 Hz in those studies. 3.2. Fretting of base metals Figure 4 shows the contact resistances in the fretting of all-base metal systems, copper against copper and nickel against electroplated nickel at the standard conditions of 4 Hz, 50 gf and tracks 20 pm long. Be-Cu alloy against itself and BeCu against copper gave curves identical with copper on copper; these are not shown. Figure 5 gives the results for solder plate on itself. The data plots are from runs which illustrate typical results. In most cases, several runs were made with each material combination, and similar behaviors were obtained. Results with gold contacts, included in these figures, are discussed later. In all cases the contact resistance falls from its initial value as the superficial oxide is worn away. However, the contact resistance then rises abruptly, although erratically, after about 20 - 200 cycles and attains values in excess of 1 s1 within several thousand cycles. Metallic contact is lost before this level of resistance is achieved, the small remaining conduction being due to loose interface material consisting of oxides and oxide-covered wear particles which become more voluminous as fretting continues.
Fig. 4. Contact resistance us. number of fretting cycles for base and for solid gold riders against base metal flats (standard conditions: 4 Hz; 50 gf; 20 pm tracks). 1003
10: C ?
Material Rider -L!!
and Thickness (pm) flaJ “nd.rpl.te
Fig. 5. Contact resistance vs. numbq and solder-plated members.
of fretting cycles for various combinations
3.3. Fretting of solid gold against gold-plated copper Runs were made with solid gold riders and Co-Au-plated copper flats of roughness 0.4 pm center-line average (c.1.a.). There was a pronounced tendency for macroscopic welding of the specimens, which could be de-
_....., . ,I,,1 105
Fig. 6. Contact resistance us. number of fretting cycles for solid gold riders against 0.05 pm Co-Au-plated copper flats, with and without a nickel underplate. 1
1 , .,,,,., ,,,,,,, ,,,,,, . ,.,,,_,,,,_, 1 1 1,.,,,1 100
103 [email protected]
Fig. 7. Effect of surface roughness on contact resistance us. number of fretting cycles for solid gold riders against 0.13 pm Co-Au-plated copper flats.
tected by a fall in contact resistance to below 0.3 mS2 after several thousand cycles. Coefficients of adhesion of from 1 to more than 10 were obtained. The thicker deposits were more prone to weld. The results where welding did not occur are presented in Fig. 4 for unplated copper, Fig. 6 for a 0.05 I.cm deposit, Fig. 7 for a 0.13 E.trn deposit, Fig. $ for a 0.23 pm deposit and Fig. 9 for a 0.6 pm deposit. All runs show similar behavior: a period of low and relatively constant contact resistance for several hundred to as many as 2000 cycles, which is followed by a rise to a peak after lo4 - lo5 cycles and then a fall and leveling at a value somewhat higher than the initial values. The curves for unplated copper and gold-plated copper at thicknesses up to 0.23 pm are remarkably similar. The run with gold 0.6 E.trn thick shows attenuation of the change in contact resistance with a peak value of 4 ma. Overall, the stability of the contact resistance improves in the systems from copper against copper to solid gold against 0 - 0.23 pm gold-plated
Fig. 8. Contact resistance us. number of fretting cycles for solid gold riders against 0.23 I.trnCo-Au-plated copper flats, with and without a polyphenyl ether contact lubricant.
1 100 ,
. . ,,,,”
, , ,,,,.,
Fig. 9. Contact resistance us. number of fretting cycles for solid gold riders against 0.6 pm C-Au-plated copper flats, with and without a nickel underplate.
copper to solid gold against 0.6 pm gold plate. Presumably, runs with flats having Co-Au deposits thicker than 0.6 pm would show a greater stability of contact resistance. Runs were made with solid gold against 0.13 E.tmgold plate on copper that was randomly abraded to 0.1 or 0.9 pm c.1.a. prior to plating for comparison with the standard finish of roughness 0.4 I.trnc.1.a. The results, given in Fig. 7, are similar for all roughnesses. In contrast, base metals have been found to wear more when smooth than when rough, as discussed in ref. 7, p. 122; for coatings with thicknesses that are of the same order as the scale of surface roughness, the less the roughness, the greater is the actual thickness of gold which compensates for what presumably would be a greater wear rate. The reason that similar gold deposits are thicker on smooth surfaces originates in the technique for determining thickness; a gravimetric method was used (Section 2.1.2) in which it was assumed that the surfaces are ideally smooth, whereas the same mass of gold is actually spread thinner
on rough surfaces since their area is larger. The dependence of the sliding wear of gold platings on roughness over long tracks has been explained in this way  . The amounts of fretting of gold both thicker and thinner than the 0.13 pm deposits on 0.4 pm c.1.a. surfaces were similar, which further emphasizes the insensitivity of fretting to topography in these experiments. 3.4. Fretting of solid gold against plated gold with a nickel underplate on a copper substrate Figures 4,6 and 9 show the results from runs with solid gold riders against 2.5 pm nickel plate on copper, nickel plate with a top layer of 0.05 pm Co-Au and nickel plate with a top layer of 0.6 I.crn Co-Au respectively. Data for comparable runs without nickel plate are also given in these figures. It is clear that nickel is highly beneficial in reducing the contact resistance instability in fretting with solid gold riders, although the all-nickel system is very poor (Fig. 4). Runs with gold riders against nickel-coated flats do not give contact resistances greater than 3.5 ma. As discussed later, this is due to the low wear rate of nickel and the marked tendency of pure gold to transfer to it; this effectively makes the system all gold. Indeed, a larger fraction of runs attempted with nickel failed by cold welding to the solid gold rider than when the nickel plate was omitted. With Co-Au electroplate on nickel, as little as 0.05 pm gold is able to keep the contact resistance from rising over 1 ma beyond 1000 cycles. Undoubtedly, the underplate lowers the wear rate of the gold, as found previously  in sliding on long tracks. The role of the thin gold deposit on the flat can be viewed as protection of the base metal from oxidation until significant transfer horn the gold rider has occurred. 3.5. Fretting of gold against 60Sn-40Pb solder plate Figure 5 shows the results from the sliding of solid gold riders against 25 pm solder plate on copper and for solder-plated riders against 0.6 pm CoAu on a nickel underplate over copper. Data for the a&solder system are included for comparison. Overall, results in the Au-solder system are poor. Unlike gold against copper and especially gold against nickel (Fig. 4), where the contact resistances did not exceed 25 ma and 3.5 mS2 respectively, the contact resistance with gold against solder attains values of 1 - 100 a ; this would be unacceptable in virtually all circuit applications. As shown later, this result is due to the rapid transfer of solder to gold to make a solder-solder contact. The relative hardnesses of the contact materials mainly control the net direction of transfer, solder being considerably softer than gold. 3.6. Fretting with 7OAu-30Ag alloy riders Because of the apparent dependence of the contact resistance stability in hetting on the relative hardnesses of the solid noble metal and of the opposing contact member, whether the latter is base metal or plated with
Fig. 10. Contact resistance vs. number of fretting cycles for solid 70Au-30Ag against base metal flats.
Fig. 11. Contact resistance vs. number of fretting cycles for solid 70Au-30Ag alloy riders against 0.05 pm Co-Au-plated copper flats, with and without a nickel underplate.
Co-Au, runs were conducted with riders made of 70Au-30Ag alloy which is harder than pure gold (Table 1). Figure 10 shows the results from fretting against copper and nickelplated copper flats. Contact resistance peaks in excess of 0.1 fi can be seen; this is about 10 - 100 times greater than the maximum values with solid pure gold riders (Fig. 4). The shapes of the plots of contact resistance against the number of cycles are the same for both solid noble metals. Figures 11 and 12 show the results from runs with 70Au-30Ag alloy riders and 0.05 and 0.6 pm Co-Au flats, both with and without a nickel underplate. The Co-Au plating generally improves the contact resistance stability compared with the results with unplated flats, either by delaying the onset of significant increases in contact resistance or by attenuating its peak value. The nickel underplate appears to be beneficial, except in the runs where beyond lo5 cycles its use was accompanied by large excursions of contact resistance.
Q&S Fig. 12. Contact resistance us. number of fretting cycles for solid 70Au-30Ag alloy riders against 0.6 pm Co-Au-plated copper flats, with and without a nickel underplate.
By comparison of the results with those for gold riders (Figs. 10 and 4; Figs. 11 and 6; Figs. 12 and 9), it is clear that 7OAu-30Ag alloy is inferior to pure gold. However, there is no macroscopic welding of the 70Au-30Ag alloy to the flats, unlike pure gold; this supports the view that, because this harder metal has less tendency to transfer by adhesion, it has a poorer behavior than pure gold. 3.7. Effect of lubrication on fretting It has been found that lubricants are able to stabilize the contact resistance of base metal against base metal pairs during fretting [2,10,11] . It was therefore of interest to determine whether a contact lubricant affects the behavior of noble metals. The initial experiment involved solid gold riders on unplated base metals, copper and nickel-plated copper, as shown in Fig. 13. The results are not very different from those with unlubricated contacts shown in Fig. 4. This suggests that, although the polyphenyl ether is a good boundary lubricant for most metals, any reduction in the rate of generation of wear debris during fretting may have been counterbalanced by its ability to inhibit the transfer of gold from the rider to the flat. Runs with typical all-gold systems, with lubricated flats which have 0.13 and 0.23 pm of Co-Au plate on copper or with 0.6 Mm of Co-Au on a nickel underplate on copper were superior to corresponding runs without a lubricant, as shown in Figs. 14,8 and 15, using pure gold riders. The advantages of runs at a higher load, 100 gf compared with 50 gf, are shown in Fig. 8. The contact resistance in lubricated fretting which steadily falls to low values with 0.6 pm Co-Au plate (Fig. 15) suggests that the area of
Fig. 13. Contact resistance us. number of fretting cycles for solid gold riders against base metal flats lubricated with polyphenyl ether.
Fig. 14. Contact resistance vs. number of fretting cycles for solid gold riders against 0.13 pm Co-Au-plated copper, with and without a polyphenyl ether lubricant.
Fig. 15. Contact resistance vs. number of fretting cycles for solid gold riders against 0.6 pm Co-Au with a nickel underplate on copper, with and without a poiyphenyl ether lubricant.
Fig. 16. Contact resistance us. number of fretting cycles for solid 70Au-30Ag alloy riders against lubricated 0.6 Pm Co-Au-plated copper, with and without a nickel underplate.
metallic contact may grow steadily because of burnishing, as found previously  in runs with gold on long tracks. Additional runs were made with riders of ‘7OAu-30Ag alloy using flats of 0.6 ,um Co-Au on copper, with and without a nickel underplate (Fig. 16). By comparing these results with those from unlubricated sliding in Fig. 12, it is evident that there is an improvement similar to that observed with pure gold riders. With lubrication, no runs were aborted because of the cold welding of either the pure gold or the ‘70Au-30Ag alloy riders. As shown later, the benefits of lubricant in all-noble metal systems are due to its ability to lower the wear rate of gold and thus to postpone the onset of rising contact resistance which occurs when the base metals exposed by wear oxidize. Although lubricants have been postulated [2,7] to inhibit the oxidation of base metals, there is no evidence that this is an essential feature in the ability of polyphenyl ether to maintain a low contact resistance in the systems examined. 3.8. Effect of aging of samples prior to fretting Connector contacts sometimes acquire surface films due to aging. It was therefore of interest to determine the effect that this condition would have on the contact resistance during fretting. ‘Iwo treatments were used: (a) thermal aging in air for 1032 h at 125 “C with flats made of copper, nickel plate on copper and Co-Au plate on copper with and without a nickel underplate; (b) exposure of copper and Co-Au-plated copper flats to a tarnishing atmosphere of flowers of sulfur in a sealed glass vessel at 85% relative humidity and 50 “C for 90 h. Insulating films develop on gold by the thermal diffusion of the codeposited cobalt and of substrate or underplate metals to the surface where they oxidize  . With sulfur, tarnishing occurs by reaction at pores in the deposit, followed by spreading over the
Fig. 17. Scanning electron micrographs of worn flats from fretting (4 Hz; 50 gf; wipe distance, 20 pm; solid gold riders): (a) nickel-plated copper after 300 000 cycles; (b) 0.05 pm Co-Au on a nickel underplate on copper. The specimens were cold welded (coefficient of adhesion, 2.3).
gold surface of the copper sulfides and oxides which result. Tests such as these [ 121 are commonly used to evaluate the connector performance. By determining the contact resistance for a 50 gf load of the samples with a probe [ 131 identical with the solid gold rider, it was found that there were significant increases from the values with unaged samples for Co-Au plated layers at thicknesses up to 0.13 pm. Fretting experiments with gold riders were conducted on the samples, both with and without lubrication. The results are similar to those already described without aging, except that the initial values of the contact resistance are high but diminish rapidly with fretting to values identical with those for the comparison samples. Also, the runs have less tendency to fail because of the cold welding of rider and flat. These results indicate that the fretting of Co-Au-plated flats is unaffected by prior aging. However, fretting corrosion may depend on the atmosphere (as discussed in ref. 7, p. 124), since corrosion processes are environmentally dependent. 4. Analysis and discussion 4.1. Examination of worn surfaces The varied contact resistance behaviors that result from fretting (described above) can be understood from the examination of worn riders and flats. Optical microscopy is a particularly useful technique since it enables us to identify large particles and areas of gold, nickel or copper easily by their characteristic colors. Scanning electron microscopy with energydispersive X-ray analysis (EDXA) was used to clarify the fine details of the surface structures and compositions.
Fig, 18. Scanning electron micrographs of worn flats from runs of increasing duration with solid gold riders on 0.05 pm Co-Au-plated copper flats: (a) lo3 cycles; (b) lo* cycles; (c) 105. cycles;(d) lo6 cycles.
4.1.1. Solid gold riders against nickel-plated flats
Figure 17 shows specimens from runs with gold riders against nickel plate or Co-Au on nickel-plated copper flats after 300 000 cycles. Figure 17(a), for a flat which had no Co-Au plate at the beginning, displays extensive transfer of the rider metal; 50% of its surface is covered with gold. The nickel plate is not worn very much because the nearly continuous original surface scratches are clearly visible. The low contact resistance, 1.4 mSt (see Fig. 4, lowest curve), confirms that contact is chiefly metallic, gold on gold. Wear debris is mainly gold, containing only a small amount of nickel, and probably was formed in a second stage, after transfer. As pointed out earlier, the pronounced tendency of pure gold to transfer to much harder nickel can cause macroscopic welding, and Fig. 17(b) shows the worn flat from such a run where the coefficient of adhesion was 2.3. The fretted region is nearly all transferred gold. 4.1.2. Solid gold riders against copper flats Figure 18 shows worn flats from typical runs with gold riders against Co-Au-plated copper from runs of increasing duration. The flats appear to be similar for Co-Au platings in thicknesses from 0 - 0.23 Frn. The flat in
Fig. 18(a) after lo3 fretting cycles (contact resistance, 0.45 ma ) is lightly burnished. Although the surfaces were rigorously cleaned prior to test, traces of adventitious contamination provide surface protection. This condition is identical with that found previously [ 141 in the unlubricated sliding of gold on long tracks where the contaminant came from the environment during running. This protective material is evident as dark patches on the worn surface in the scanning electron micrograph and is presumed to be organic. Figure 18(b) shows a surface after lo* cycles (contact resistance, 6.0 ma ). The rider and the flat have identical appearances. Adhesive transfer is occurring and the surface scratches of the unworn contact have been obliterated. The few loose particles are large and contain both copper and gold. Figure 18(c) is a result of continuing the analysis with a run of lo5 cycles (contact resistance, 3.3 mQ ). The flat is smoother than that in Fig. 18(b); this means that the extent of adhesive transfer has diminished. Gold still covers the surface in patches. Figure 18(d) shows a worn surface after very long sliding, lo6 cycles (contact resistance, 3.0 rnfi ). The rider and flat have identical appearances. Optical microscopy shows that less than 10% of the surfaces is covered with gold. The surfaces are relatively smooth and probably are a mixture of gold, copper and copper oxides. Copious loose debris has formed. Figure 19 shows the worn surface of Fig. 18(d) in greater detail. Figures 19(a) and 19(b) show different areas from the central region of the flat. The small particles of debris contain both gold and copper according to an EDXA. Figure 19(c) shows the upper left quadrant of Fig. 19(b) at a higher magnification. Surface cracks suggest that delamination wear is occurring [ 151, i.e. the repeated cyclical stresses of fretting cause subsurface cracks to form and propagate to the surface with resultant loosening of the sheets of metal. However, loose debris (Fig. 19(d)) is not predominantly sheet like, which implies that much of it is the result of the loosening of particles that had been transferred between members earlier. It is likely that fretting with adhesive transfer of gold between both members persists until the gold plate has been worn away. There may then be a short interval during which adhesive transfer continues, with metallic copper from the flat now being involved. Oxide soon builds up, however, during which the rate of adhesive transfer diminishes to zero. Delamination wear would be expected to become increasingly significant during this phase. Relatively few loose particles form during adhesion, while much more debris is generated in the postadhesion stage. The changes in contact resistance are consistent with this mechanism; significant rises occur only when the surfaces become extensively covered with insulating oxide once the gold plate has worn away. Delamination wear has been found [ 151 to be important in the fretting of steels, and it occurs after an earlier stage in which adhesive transfer and wear predominate. When adhesion changed to delamination, the surfaces were found to become smoother. Observations from the present study are similar to those from all-ferrous systems, with the added complexity that the gold plate must first wear off before oxide forms with degradation of the contact resistance.
Fig. 19. Scanning electron micrographs of an ~uilibrium worn surface of a flat (Fig. 18(d) in greater detail): (a), (b) central region of the scar; (c) cracks believed to be due to delamination wear; (d) wear debris.
42.3, 7OAw3OAg alloy riders against nickel-plated flats Figure 20 shows the rider and flat from a run (Fig. 11) with a 7OAu30Ag ahoy rider on a Bat with 0.05 pm Co-Au on a nickel underplate on a copper substrate. The worn areas have identical appearances and compositions. Only a few flecks of gold are on the surfaces. The nickel underplate is severely worn, and the coverage of the rider and flat by nickel is virtually complete. Unlike the companion run with a pure gold rider (Fig. 17), when 7OAu-30Ag ahoy is used, nearly ah-base metal surfaces appear in prolonged sliding. After 300 000 cycles the contact resistance is 140 mS2. The wear debris is mostly nickel, with a few particles of gold. 4.1.4. Lubricated fretting Figure 21 illustrates worn surfaces after lo6 cycles from a run with 0.6 I.tmCo-Au on a nickel under-plate on copper. The contact resistance stability (0.3 mSI at the end of test) is superior to that of any other material combination in this study, as illustrated in Fig, 15, because a relatively thick Co-
Fig. 20. Scanning electron micrographs of worn specimens for a 7OAu-30Ag alloy rider against 0.05 pm Co-Au on a nickel underplate on copper (300 000 cycles; contact resistance, 140 ma). The gold on the flat has worn through and the rider is coated with nickel by transfer.
Au-plated layer on an underplate which improves the wear resistance of the gold plate was used with a solid gold mating member. Nevertheless, the CoAu deposit is worn through in the center of the fretted region; 25% of the area is nickel, as shown in the optical photomicrograph (Fig. 21(a)) and the corresponding scanning electron micrograph (Fig. 21(b)). These photographs were obtained after the surfaces had been tho~ughly washed in l,l,l-trichloroethane, the diluent from which the lubricant was originally applied. Figure 21(c) shows dark material surrounding the wear track, and Fig. 21(d) at a higher magnification reveals that it has a gelatinous appearance with embedded gold particles (EDXA). It is probable that this substance is friction polymer , for a paste of wear debris in oil would have been removed readily by the solvent wash prior to scanning electron microscopy examination. It has been found [ 16,171 that pure gold can generate polymer after prolonged wiping or sliding at a low load in benzene vapor which, while able to lubricate, still permits sufficient Au-Au asperity contact to give a low contact resistance.
Fig. 21. A worn flat from a run with a solid gold rider against a polyphenyl-ether-lubricated copper flat which has 0.6 pm Co-Au on a nickel underplate ( lo6 cycles): (a) overall view with optical microscope; (b) overall view with scanning electron microscope; (c) the friction polymer surrounding the fretted area; (d) same as (c) at a higher magnification. The embedded gold wear debris should be noted.
In this work which involves dissimilar contact materials fretted in air, the direction and extent of metal transfer were found ultima~ly to control the contact resistance. A clean all-solid gold contact can be considered to be a limiting case in which massive cold welding tends to occur, particularly as the load is increased. A clean all-solid base metal contact, such as copper on copper, is another limiting system in which the contact resistance rises quickly to high values as oxide wear debris accumula~s between the surfaces. When one member is clean solid pure gold and the other is a base metal and provided that gold transfers readily to the base metal, the contact resistance does not rise significantly; an example is pure gold against nickel. If base metal transfers to the gold contact, the surface becomes all base and the contact resistance rises to high levels, as with gold against solder plate. At equilibrium, both surfaces have the same compositions and structures. Another example is 70Au-30Ag alloy against nickel plate on copper. Because 7OAu-30Ag alloy has a greater hardness than solid pure gold, systems
Fig. 22. A schematic diagram of the contact resistance trends during a run with a solid gold rider against Co-Au-plated copper.
with the ahoy have an inferior contact resistance; the direction of adhesive transfer is predominantly from the base to the noble metal. For solid pure gold against copper the metal transfer of each member occurs approximately to the same extent. This is not surprising, since the materials have virtually identical structures and mechanical properties. However, copper is base and, after prolonged fretting, the resulting surfaces (which consist of gold, copper and copper oxides) display a contact resistance intermediate between that of gold against gold and copper against copper. The contact resistance actually is close to that of an all-gold system because relatively little metallic contact is necessary for a low contact resistance. From this model the effect that the variation in surface composition of the flat and that lubrication have can be explained. A gold electrodeposit on base metal substrates makes the system all gold, and a low contact resistance is maintained until the gold layer is worn away. The thicker the gold plate, the more fretting cycles can occur before the contact resistance degrades. However, a difference in the gold thicknesses when the deposits are relatively thin, e.g. in the approximate range 0.05 - 0.23 pm, may not be significant, since such platings appear to wear away quickly. Lubrication effectively maintains a low contact resistance with contacts involving gold or 70Au30Ag alloy against Co-Au-plated base metal because it reduces the rate of wear of gold plate on the base member, thus postponing the appearance of insulating oxides. Lubrication also reduces the tendency of gold contacts to massive cold welding by inhibiting the fo~ation of adhesive bonds. 4.3. Stages in the change of contact resistance with fretting: pure gold
against copper-based contacts Solid pure gold against unlubricated copper or Co-Au-plated copper has a complex contact resistance behavior, involving a rising then falling contact resistance (Figs. 4,6,7,8 and 14). The rise is attenuated with a thick (0.6 iurn) Co-Au plated layer (Fig. 9). From the analysis in Section 4.12 the contact resistance trends can be related to the transfer and wear which occur. Figure 22 schematically illustrates the contact resistance behavior and is divided into three regions. Region 1, typically several thousand cycles in
duration, involves a burnishing of the surfaces with negligible transfer and wear. If the surfaces have a thick insulating film, as with the thermally aged or sulfided specimens of this study, a preliminary stage (Pre 1) of sharply falling contact resistance occurs as the film wears off during fretting. Region 2, from less than lo4 to more than lo5 cycles, involves severe metallic transfer; this is further subdivided into region 2a with rising contact resistance and then region 2b during which contact resistance falls. In region 2a, Co-Au plate, if present, wears out. Meanwhile the copper and copper oxide which appear in its place cause the contact resistance to rise. The smoothing of the contacts in region 2b can be postulated as the reason for the fall in contact resistance because, as with the burnishing of region 1, the contact area grows. Finally, by region 3 where the contact resistance stabilizes, although at a level higher than the initial value, the equilibrium surface consisting of mixed gold, copper and copper oxide has developed. Both the rider and the flat have identical surface compositions. Overall, regions 1 and 2 are the wearing-in stages , and region 3 is the equilibrium fretting stage, which is expected to continue indefinitely. It is likely that the same stages in the contact resistance trends occur at other operational conditions with these contact materials, modified only in the level of contact resistance and the numbers of cycles where the transitions occur as the load, the cycle rate and other fretting parameters are varied.
The outstanding fact to emerge from this study is the ability of contacts to maintain a relatively low and stable contact resistance during fretting when one member is solid pure gold. The chief requirement in this case is that the opposing base metal contact (whether or not it has a gold plating before fretting wear) does not transfer to the solid gold member, thus making the system all base. Nickel plating is particularly desirable because pure gold readily transfers to it. Also a nickel underplate enhances the durability of a thin gold deposit. Noble metals harder than solid pure gold, such as solid 70Au-30Ag alloy, transfer less readily, and the contact interface is therefore more likely to have a significant base metal coverage eventually with a resultant increase in the contact resistance. Solid pure gold against Co-Auplated copper is an intermediate case where the surface coverage at equilibrium entails a mixture of gold and copper-based materials which have a higher contact resistance than when the underplate is omitted. The contact resistance rises and then falls to an equilibrium level, which can be explained from an analysis of the surface coverage and topography at various stages during a fretting run. In order to maintain a low contact resistance in a mated system, the most effective practice is to avoid fretting movements entirely, by appropriate design or usage of the connector. Design expedients include increasing the normal load and careful selection of the construction materials of the
connector so as to minimize the dimensional changes during thermal fluctuations if these could result in fretting movements. When fretting cannot be avoided, thicker gold finishes are obviously desirable, but a nickel underplate may be adequate to provide the required stability of contact resistance. A contact lubricant is also helpful, since it can lower the wear rate, thus extending the life of thin gold platings. Acknowledgments E. T. Ratliff plated the gold- and nickel-coated specimens used in this study. E. S. Sproles and F. E, Bader reviewed the manuscript and provided useful comments. References 1 M. Antler, Connectors, 2
4 5 6
7 8 9 10
14 15 16 17
in F. H. Reid and W. Goldie (eds.), Gold Plating Technology, Electrochemical Publications, Ayr, 1974, Chap. 36, p. 490. E. M. Bock and J. H. Whitley, Fretting corrosion in electrical contacts, in fioc. Holm Co& on Electrical Contacts, Illinois Institute of Technology, Chicago, IL, 1974, pp. 128 - 138. F. E. Bader, S. P. Sharma and M. Feder, Atmospheric testing of connectors - a new accelerated test concept, in Proc. 9th Int. Conf. on Electric Contact Phenomena, Illinois Institute of Technology, Chicago, IL, 1978, pp. 341 - 351. ASTM Stand. B539, 1970. M. Antler, The lubrication of gold, Wear, 6 (1963) 44 - 65. D. P. Schnorr, Design and application of rigid and flexible printed wiring, in C. A. Harper (ed.), Handbook of Electronic Packaging, McGraw-Hill, New York, 1969, Chap. 1, pp. 1 - 43. R. B. Waterhouse, Fretting Corrosion, Pergamon, New York, 1972. M. Antler, Wear of gold contact finishes: the importance of topography, underplate, and lubricants, Znsul. Circuits, 26 (1) (1980) 15 - 19. M. Antler and M. H. Drozdowicz, Wear of gold electrodeposits: effect of substrate and of nickel underplate, Bell Syst. Tech. J., 58 (2) (1979) 323 - 349. W. 0. Freitag, Wear fretting, and the role of lubricants in edge card connectors, in Proc. Holm Conf. on Electrical Contacts, Illinois Institute of Technology, Chicago, IL, 1975, pp. 17 - 23.. W. H. Abbott and J. H. Whitley, The lubrication and environmental protection of alternatives to gold for electronic connectors, in Proc. Holm Conf. on Electrical Contacts, Illinois Institute of Technology, Chicago, IL, 1975, pp. 9 - 16. M. Antler, Contact properties, in F. H. Reid and W. Goldie (eds.), Gold Plating Technology, Electrochemical Publications, Ayr, 1974, Chap. 22, pp. 277 - 294. M. Antler, Contact resistance probing: development of a standard practice by the American Society for Testing and Materials, in Proc. 10th Int. Conf. on Electrical Contact Phenomena, Budapest, August 25 - 29, 1980, Scientific Society for Telecommunication, Budapest, 1980, pp. 13 - 21. M. Antler, Wear of gold plate: effect of surface films and polymer codeposits, IEEE Trans. Parts, Hybrids, Packag., IO (1974) 11 - 17. R. B. Waterhouse, The role of adhesion and delamination in the fretting wear of metallic materials, Wear, 45 (1977) 355 - 364. H. W. Hermance and T. F. Egan, Organic deposits on precious metal contacts, Bell Syst. Tech. J., 37 (1958) 739 - 776. W. H. Abbott and W. E. Campbell, Frictional polymer formation on precious metals experimental observations, in Proc. 9th Int. Conf. on Electrical Contact Phenomena, Illinois Institute of Technology, Chicago, IL, 1978, pp. 359 - 362.